Drawing the lines around
VoIP isn't so easy
Fred Goldstein, January 2004
...Modern VoIP switches are relatively inexpensive and flexible, and raw bandwidth is cheap, so it can make sense to use VoIP...
.
...regulation and technology are inseparable. Regulation should respond to technology; technology will respond to regulation!...
...The logic of taking away common carrier status from DSL is not that far removed from the logic of taking away common carriage status from any telephone call tained along the way with IP ...
...granting preference to VoIP compared to other technologies is a dangerous move as well, especially since there's no unambiguous way to tell which is which...
...any system of arbitrary call classification is problematic. VoIP is the canary in the coal mine...
Conventional wisdom within the competitive-telecom community is that Voice over Internet Protocol should be unregulated. Sometimes, VoIP advocates call for VoIP to remain unregulated, which assumes that it currently is unregulated, and that its unregulated status is threatened. Indeed there are some threats to its status, as state utility regulators cry foul at companies like Vonage who walk and talk like telephone companies, but deny that they really are. In response to this confusion, the FCC plans to make some kind of VoIP ruling later this year. The problem, however, is not a simple case of regulated or unregulated. Before that decision can be made, it helps to know what one is referring to. VoIP, it seems, occurs in many different flavors, and it is neither simple nor, I suggest, rational, to carve out a special regulatory status for a telephone service just because it uses IP.
Nobody wants to touch computer-to-computer voice
Well, maybe not "nobody", but realistically, there's little controversy about voice being sent between two computers. Microsoft's NetMeeting has been supporting this for years, and the original ITU-T Recommendation H.323, the first "standard" for VoIP, was based on NetMeeting. True, when VocalTec introduced its VoIP software in the mid-1990s, a trade group of small long distance resellers, America's Carriers Telecommunications Association (ACTA), filed an FCC petition asking for such software to be regulated as if software were itself a form of common carriage. But ACTA harmed itself by going overboard, becoming in its own small way the SCO of its day. (ACTA itself faded away, merging into the Competitive Telecommunications Association in 1999.) If a voice connection doesn't touch the public switched telephone network (PSTN), it's beyond the scope of telephone regulation. The PSTN is regulated; talking isn't. I don't think that's even controversial any more.Sometimes VoIP is used transparently within regulated networks
The opposite extreme occurs when Voice over IP is used behind the scenes, as an implementation tool. One example of this occurs when long distance carriers use VoIP on their intercity trunks. Several major carriers do this, to some extent, typically with uncompresed voice on dedicated trunk facilities. Modern VoIP switches are relatively inexpensive and flexible, and raw bandwidth is cheap, so it can make sense to use VoIP (at about 90-100 kbit/second per call, vs. 64 kbit/second for traditional TDM) on some interswitch trunk lines. Voice quality is preserved, though there may be a few milliseconds of extra latency compared to TDM. These carriers pay the same access charges as everyone else, and don't deny that they're in the "telecom" business. Local exchange carrier too are finding additional applications for VoIP, again mostly transparent, both for interswitch trunking and for connecting remote terminals (digital loop carrier) to host switches. The Occam "Broadband Loop Carrier" is an example of a device that allows local network operators to transparently use VoIP in lieu of TDM.Another example of transparent VoIP occurs in PacketCable, the CableLabs standard for providing telephone service across a cable modem network. PacketCable makes use of high-priority bandwidth on compliant (DOCSIS 1.1 or later) cable modems to provide full-quality local telephone service. It's normally provide by the cable company itself, which is interconnected, as a CLEC, to the rest of the telephone network. Again, the fact that there is IP encapsulation within the PacketCable network is transparent to the user. While CLEC status incurs some costs, it also has significant benefits, so most cable operators should have no problem with it.
But according to some VoIP advocates, the mere use of IP anywhere within a phone call should automagically transform it from "telecommunications" into "information", a crucial regulatory distinction. It makes a nice sound bite, but is it reasonable? Should the mere presence of an IP header on a wire somewhere qualify a network for preferential treatment?
Regulation has to line up with technology
It's important to note again (this is, after all, ionary's raison d'etre) that regulation and
technology are inseparable. Regulation should respond to technology;
technology will respond to
regulation! When I work with a new network provider, I have to
take into account local regulatory conditions. What works in
Chicago might not work in a small market with a non-Bell incumbent
LEC. Regulations should not assume, then, that every potential
network architecture has already been developed. If regulators
create a distinction that can be met with a technological solution, even
one that doesn't exist today, then someone will create it.Phone to what?
So what are the different specific details that a regulated-or-not decision for some particular instance of VoIP will have to face? An initial breakdown is to look at how the PSTN interacts with the netework, because it's usually not VoIP that's regulated, it's the network that VoIP connects to! In its 1999 Report to Congress, which was itself intended to forestall heavy-handed pressure against VoIP, the FCC suggested that "phone-to-phone" VoIP calls were, most likely, telecommunications. This suggested (but did not make a final rule) that other VoIP calls were not.But earlier rulings, which to be sure predate the Telecom Act and VoIP, have suggested that when a telephone call is carried across a state line and then meets a local exchange network, the local exchange network is entitled to charge switchedaccess (the subsidy-laden rate that long distance carriers pay) for what is in effect a leg of a long-distance call. That rule has been applied to the open end of dedicated access facilities, such as calls carried over tie lines between PBX systems before being sent into the local network ("tail end hop off"). So that begs the question: If VoIP is not subject to access, can a company get around paying access rates on tail end hop off by converting its tie lines into VoIP? Can a long-distance carrier do the same on its dedicated-access ("WATS-type") services? Does anyone remember Execunet, the pioneering MCI phone-to-phone discount long distance service whose controversial creation led to the introduction of switched access charges? Disco-era regulations are having a hard time with today's technology.
Vonage escaped state regulation in Minnesota because, the judge noted, it is not a phone-to-phone service. It requires a dedicated computing device (such as an ATA-186) at the customer site. Thus under the FCC's tentative guideline, it's not a "telecommunications service". So for now, "computer to phone" calls may qualify as "information", and the open end of the circuit, where the VoIP provider (noting that these companies often deny being a "carrier") hands off calls to a local exchange carrier, is probably eligible for local, rather than access, rates. One just might consider this to be regulatory arbitrage. But it's not a huge threat to "the sysem", either, because it only works if the subscriber has a broadband connection.
A continuum of options
One common rationale for granting VoIP certain privileges seems to be that because IP is normally used for information transmission, not voice telecommunications, then anything carried over it is really information, not merely telecommunications. The medium is the message, and the message is the medium. That's a fallacy, of course, and it cuts both ways. ILECs are indeed trying to take away independent ISP access to their DSL services by claiming that their DSL telecom services are really indistinguishable from their captive-ISP information services, and thus shouldn't be tariffed common carriage. (Existence proof of third-party ISPs being, of course, a detail to ignore.) The logic of taking away common carrier status from DSL is not that far removed from the logic of taking away common carriage status from any telephone call tainted along the way with IP. One is pro-competitive and one anti-competitive, of course, but these are the consequences of ill-thought-out decision-making. (If the ILECs get their way and shed ISP competitors from their networks, will their unregulated information services attempt to block parasitic VoIP traffic such as Vonage?)Another ratioanle is that VoIP does not transparently pass the voice, but processes and degrades it, at least at the IP header layer. There is some factual basis here when applied to low-bit-rate-voice discount calling services, such as those that even work over dialup. But should degraded service be privileged? How degraded does it have to be?
Let's assume that VoIP is given preferential treatment per se. There are many theoretical ways to turn TDM voice into VoIP. What provider would not use one of them, in exchange for lower call origination or termination costs? Here are some examples. Is there a bright line? Which of these ways of carrying voice should be considered "information" and which are merely "telecom"?
- Uncompressed TDM voice is placed into packets, a few
milliseconds at a time, with compressed 2-to-4-byte packet headers,
carried over a dedicated (voice-only) lossless network.
- As above, but what if the packet headers are called "Frame Relay" or "ATM", not "compressed IP"?
- Uncompressed voice is time division multiplexed on a physical pipe with time slots that contain data.
- Uncompressed voice is time
division multiplexed on a physical pipe, but gratuitous IP
headers are regularly placed into the voice time slots.
- Uncompressed TDM voice is placed into packets, a few milliseconds at a time, with uncompressed IP headers, carried over a dedicated lossless network.
- TDM voice is placed into packets a few milliseconds at a time, compressed to 32 kbit/second ADPCM if a modem tone is not detected, with "compressed IP" 2-to-4-byte packet headers, carried over a dedicated lossless network.
- TDM voice is placed into packets 20-40 milliseconds at a time, compressed to a low bit rate with a vocoder such as G.729 or G.723.1, "compressed IP" 2-to-4-byte packet headers, carried over a dedicated lossless network.
- What if the ADPCM or low-bit-rate voice used uncompressed IP headers instead?
- What if the dedicated lossless VoIP network also carried,
as 1% of its traffic, some IP data packets from an affiliated ISP?
What if data were 20% of the traffic? Does a de minimis or "substantial amount"
test apply?
- What if the backbone link used MPLS to provide voice traffic with the necessary QoS, while carrying a substantial amount of data as well? Would it matter if the voice were compressed? Would it matter if the IP headers were compressed, or if the voice were converged directly over the MPLS layer without even using an IP header?
- What if a new packet protocol, not called "IP" were to emerge, which carried both voice and data with their native QoS requirements met? What if ATM, sans IP, were revived for the same purpose?
